Key points

History, definition, and classification

Purpura in association with pestilential fevers was described by Hippocrates, but it was not until the sixteenth century that purpura in the absence of fever was observed by Lusitanus. It took another 3 centuries for Krauss (1883) and Denys (1887) to recognize that purpura hemorrhagica was caused by a decrease in blood platelets. The association of drug and purpura was first reported by Vipan in 1865 before the role of platelets in hemostasis was known. Later, Rosenthal in 1928 provided evidence that drug ingestion caused thrombocytopenia by rechallenging a patient with quinine-induced thrombocytopenia with the drug and showing a prompt decrease in platelet count.

Drug-induced thrombocytopenia (DIT) can be divided into 2 categories according to the mechanism responsible for the thrombocytopenia, namely (1) suppression of platelet production and (2) increase in peripheral platelet destruction or clearance. The former, which usually occurs as pancytopenia, is caused mostly by myelosuppressive agents (mainly chemotherapeutic drugs) and their suppressive effect is dose dependent. The latter can be further divided into 3 subtypes:

• 1.
  

  Nonimmune DIT (caused by a direct toxic effect of the drug on platelets; eg, ristocetin, an antibiotic that is no longer used in clinical practice)

  
  
• 2.
  

  Immune DIT (mediated by a drug-dependent antibody)

  
  
• 3.
  

  Autoimmune DIT (mediated by a drug-independent antibody that binds and prematurely clears platelets even in the absence of the drug such as alpha-methyldopa)

Most drugs are thought to cause thrombocytopenia by a drug-dependent immune mechanism (ie, the second mechanism); this is the focus of this article and other types of DIT are not discussed.

Drug-induced immune thrombocytopenia (DITP) is a common problem in clinical practice, particularly in hospital inpatients. Patients usually develop extensive petechiae or purpura, with markedly low blood platelet levels (frequently <10 × 10 ⁹ /L), about 1 to 2 weeks after commencing a new medication. Serious bleeding including intracranial hemorrhage can occur, presenting a challenging diagnostic and management problem. Diagnosis of DITP is usually made clinically but can be difficult when there are comorbid conditions (eg, infection) and concomitant drugs that can also cause thrombocytopenia. Despite significant advances in this area of hematology, there is still a need for more research to increase understanding of the mechanisms of DITP, and to improve diagnostic tests and patient management. This article discusses the pathogenesis, diagnosis, and management of DITP, focusing on recent advances.

History, definition, and classification

Purpura in association with pestilential fevers was described by Hippocrates, but it was not until the sixteenth century that purpura in the absence of fever was observed by Lusitanus. It took another 3 centuries for Krauss (1883) and Denys (1887) to recognize that purpura hemorrhagica was caused by a decrease in blood platelets. The association of drug and purpura was first reported by Vipan in 1865 before the role of platelets in hemostasis was known. Later, Rosenthal in 1928 provided evidence that drug ingestion caused thrombocytopenia by rechallenging a patient with quinine-induced thrombocytopenia with the drug and showing a prompt decrease in platelet count.

Drug-induced thrombocytopenia (DIT) can be divided into 2 categories according to the mechanism responsible for the thrombocytopenia, namely (1) suppression of platelet production and (2) increase in peripheral platelet destruction or clearance. The former, which usually occurs as pancytopenia, is caused mostly by myelosuppressive agents (mainly chemotherapeutic drugs) and their suppressive effect is dose dependent. The latter can be further divided into 3 subtypes:

• 1.
  

  Nonimmune DIT (caused by a direct toxic effect of the drug on platelets; eg, ristocetin, an antibiotic that is no longer used in clinical practice)

  
  
• 2.
  

  Immune DIT (mediated by a drug-dependent antibody)

  
  
• 3.
  

  Autoimmune DIT (mediated by a drug-independent antibody that binds and prematurely clears platelets even in the absence of the drug such as alpha-methyldopa)

Most drugs are thought to cause thrombocytopenia by a drug-dependent immune mechanism (ie, the second mechanism); this is the focus of this article and other types of DIT are not discussed.

Drug-induced immune thrombocytopenia (DITP) is a common problem in clinical practice, particularly in hospital inpatients. Patients usually develop extensive petechiae or purpura, with markedly low blood platelet levels (frequently <10 × 10 ⁹ /L), about 1 to 2 weeks after commencing a new medication. Serious bleeding including intracranial hemorrhage can occur, presenting a challenging diagnostic and management problem. Diagnosis of DITP is usually made clinically but can be difficult when there are comorbid conditions (eg, infection) and concomitant drugs that can also cause thrombocytopenia. Despite significant advances in this area of hematology, there is still a need for more research to increase understanding of the mechanisms of DITP, and to improve diagnostic tests and patient management. This article discusses the pathogenesis, diagnosis, and management of DITP, focusing on recent advances.

Causative drugs

More than 200 drugs (including some herbal medicines) have been reported to cause DITP. Aster and Bougie identified 85 drugs that they considered as definite or probable causes of thrombocytopenia. Among these are drugs commonly used in clinical practice: heparin, quinine, antibiotics (eg, penicillin, cephalosporin, rifampicin, vancomycin, sulfonamides), antiplatelet agents, glycoprotein IIb/IIIa (GPIIb/IIIa) inhibitors (tirofiban, eptifibatide, abciximab), antirheumatic agents (gold, d -penicillamine), antiepileptics (sodium valproate, phenytoin), cardiac agents (amrinone), and the new biologic agents infliximab and rituximab. Heparin causes an immune thrombocytopenia that leads to severe thrombosis rather than bleeding, through a mechanism different from other DITPs.

Incidence

DITP incidence is not well defined and is reported to be low, ranging from 0.6 to 1.6 per 100,000. These studies used national surveillance programs and hospital records, and probably underestimate the frequency of DITP. However, the incidence of thrombocytopenia induced by GPIIb/IIIa inhibitors (tirofiban, eptifibatide, and abciximab) and heparin is higher. DITP-associated GPIIb/IIIa inhibitors have an incidence of 0.1% to 2% on first drug exposure and about 5% to 10% on subsequent exposures. This incidence suggests that thousands of patients globally must develop this drug complication every year because millions of cardiac patients receive these drugs for prevention of coronary artery thrombosis following coronary angioplasty, stent insertion, and acute coronary syndrome. Despite the variable frequencies, DITP represents a significant clinical problem because of the large number of drugs involved and the large number of patients who can be affected.

Pathophysiology

DITP is usually mediated by a drug-dependent antibody that binds to platelets or megakaryocytes (Mks) only in the presence of the sensitizing drug. Over the years, several hypotheses have been proposed for the antibody-platelet interactions. These hypotheses are discussed later ( Tables 1 and 2 ).

Hapten-induced antibody or hapten mechanism

Early immunologic studies in the 1900s suggested that small molecules like drugs are not antigenic and can only trigger an immune response when linked covalently to a macromolecule (a protein) acting as a hapten. The resulting antibodies recognize the carrier molecule only where the hapten is attached covalently. In the 1950s and 1960s, Shulman proposed the innocent-bystander or immune-complex mechanism that replaced the hapten mechanism as the dogma, and it remained accepted until the 1980s. According to the immune-complex mechanism, drug-antibody complexes bind to platelets via platelet Fc receptors. Platelets are targeted for destruction as innocent bystanders as the antibody recognizes the drug but not the platelet. It is now well recognized that the hapten and the immune-complex mechanisms are relevant for only a few drugs: high-dose penicillins cause immune hemolysis by haptens and heparin causes thrombocytopenia via immune-complex binding. These mechanisms are not discussed further in this article. Heparin-induced thrombocytopenia (HIT) represents a unique clinical-pathologic entity with strong platelet activation and frequent thrombosis. HIT merits a separate discussion with a full article to itself.

Quinine-type immune thrombocytopenia

As stated earlier, thrombocytopenia caused by quinine or its isomer, quinidine, is the prototype of most DITPs. This mechanism or a variation thereof mediates immune thrombocytopenia caused by most drugs (see Table 2 ). In 1983, we showed for the first time that, using staphylococcal protein A, we could immunoprecipitate the drug-platelet antigen (glycoprotein Ib-IX [GPIb-IX]) complex with the antibody of a patient with quinidine-induced thrombocytopenia. This radical finding contradicted the immune-complex mechanism dogma at that time, because this finding suggests that the antibody binds to platelets by its Fab but not its Fc domain. This finding was confirmed by subsequent studies from our and other laboratories. It became clear that the immune-complex mechanism does not contribute to thrombocytopenia caused by quinine, quinidine, or most other drugs, except heparin.

Current evidence suggests that the drug binds to platelet GPIb-IX (von Willebrand factor receptor complex) and/or GPIIb/IIIa (fibrinogen receptor complex).

One hypothesis (model 1, Fig. 1 A ) is that drug binding induces a conformational change of the platelet surface protein, resulting in the emergence of a cryptic epitope or neoantigen previously unseen by the immune system. The resultant antibody recognizes and binds this epitope only in the presence of the drug. In the absence of the drug, this epitope or neoantigen is buried in the glycoprotein folds or does not exist. Aster and colleagues recently proposed a new model (model 2, see Fig. 1 B) that is slightly different from the model discussed earlier. The essential difference is that, unlike model 1 in which the drug and the antibody bind the glycoprotein at different sites, in model 2 the drug and the antibody bind the target platelet glycoprotein at the same site (the drug is considered an essential component of the antibody binding site). Model 2 is supported by observations that quinine bound to platelets is more resistant to removal in the presence of the drug-dependent antibody, suggesting that the drug may be part of the binding pocket and is, therefore, entrapped by the bound antibody. Nevertheless, there is no conclusive or direct experimental evidence to support this model.

Quinine-type DITP. ( A ) Model 1. Drug binding induces a conformational change on the glycoprotein, resulting in the emergence of a cryptic epitope or neoantigen. The drug and the antibody bind at different sites. ( B ) Model 2. The drug and the antibody bind the target platelet glycoprotein at the same site. GPIX, glycoprotein IX.

Why do the drug-dependent antibodies target specific platelet glycoproteins such as GPIb-IX?

We have consistently found that all patients with quinine-induced and quinidine-induced thrombocytopenia had antibodies that reacted drug dependently with GPIb-IX ; almost all of these antibodies recognized GPIX and only an occasional antibody had specificity for GPIbα. In addition, a minority of patients (<30%) had drug-dependent antibodies reactive with GPIIb-IIIa. In contrast, other investigators found a higher proportion of anti–GPIIb-IIIa antibodies. Furthermore, we have mapped the binding site of the GPIX-specific quinine-dependent antibodies to a restricted site in the membrane-proximal ectodomain of platelet GPIX (R110–Q115) and the GPIbα-specific antibody to the N-terminal domain of GPIbα (residues 283–293).

Peterson and colleagues located the binding sites of the quinine-dependent antibodies to a 29-kDa region of GPIIIa comprising the hybrid and the plextrin-semaphorinintegrin (PSI) domains, and residues A50 and D66 played a critical role in the formation of the antibody binding sites. Patients with rifampicin-induced and ranitidine-induced thrombocytopenia also have antibodies that react with GPIX. An anti-GPIX monoclonal antibody (mAb) SZ1 cross-blocked the binding of quinine-dependent, quinidine-dependent, rifampicin-dependent, and ranitidine-dependent antibodies to GPIX, suggesting that these four antibodies bind to a site that coincides with or is close to the epitope of SZ1. It is possible that, when other less well-characterized drug-induced antibodies (eg, vancomycin-induced antibody) are studied, they may also bind to this restricted GPIX site. These data also suggest that this region must have unique characteristics that render it unusually antigenic on drug binding.

Flexible domains and immunogenicity

Protein structure studies have recently provided additional information to refine the current hypothesis. The crystal structure of GPIIIa showed that the PSI domain (which is targeted by quinine-dependent antibodies) could not be completely resolved, indicating that it possesses a degree of mobility (hyperflexibility), adopting several transient state conformations. The membrane-proximal part of the ectodomain of GPIX is also highly flexible and is likely to adopt a range of non-native or denatured states (model in Fig. 2 ). We have mapped the binding sites of anti-GPIX quinine-dependent antibodies to this hyperflexible GPIX region as described earlier.

The hyperflexible region of GPIX. ( A ) Model of human GPIX. Unstructured domains (in this case, the membrane-proximal domain of GPIX) may allow the binding of drugs of diverse chemical structures. Binding of the drug fixes a particular conformation (right panel) that, in some circumstances, is deemed foreign by the immune system. An antibody is then generated that binds to the glycoprotein in the presence of the drug (DDAb). ( B ) Hyperflexible region of GPIX binding drugs of different chemical structures.

The presence of these unstructured protein domains may allow the binding of drugs of diverse chemical structures as the protein changes from one non-native conformation to another (see Fig. 2 ). Non-native conformations are characterized by an increase in their hydrophobic surface, which may facilitate interaction by drugs containing hydrophobic pockets, such as quinine and quinidine, and it could also explain why drugs of different chemical structure, such as quinine, rifampicin, and ranitidine, interact with GPIX or GPIIIa. It is likely that each drug binds to distinct non-native structures, which then adopt a particular conformation. Under some conditions, this fixed configuration may elicit an immune response resulting in the generation of drug-dependent antibodies (see Fig. 2 ).

Thrombocytopenia induced by GPIIb/IIIa inhibitors (fiban)

GPIIb/IIIa inhibitors, tirofiban and eptifibatide (fiban), are synthetic small molecules that bind tightly to the arginine-glycine-aspartic acid (recognition sequence arginine-glycine-aspartic acid [RGD]) recognition site on GPIIb/IIIa, inhibit fibrinogen binding, block platelet aggregation, and prevent platelet thrombus formation. These drugs are widely used in clinical medicine to reduce cardiac complications following percutaneous transluminal coronary angioplasty (PTCA) and to treat acute coronary syndrome. Clinical trials and subsequent clinical experience with the fibans showed that about 0.2% and 2% of patients receiving these compounds developed an acute, frequently severe, thrombocytopenia (platelet <10 × 10 ⁹ /L) within a few hours of first drug exposure. Previous studies have shown that naturally occurring antibodies against GPIIb/IIIa cause the thrombocytopenia, thus explaining the occurrence of thrombocytopenia on first drug exposure. These antibodies are found not only in patients with fiban-induced thrombocytopenia (FIT) but also in normal volunteers never exposed to the drug. It is still uncertain whether healthy individuals with these naturally occurring antibodies will develop severe thrombocytopenia if they receive a fiban. If so, detection of the antibody before drug administration may prevent this serious complication.

Integrin research has revealed that RGD peptide and RGD mimetic compounds induce conformational changes in GPIIb/IIIa, resulting in the emergence of cryptic integrin domains that could be recognized by monoclonal antibodies specific for these ligand-induced binding sites (LIBSs). It has been suggested that antibodies from patients with FIT probably have specificity for these LIBSs. Until recently, it was uncertain whether the ligand-induced GPIIb/IIIa determinants that resulted from fibrinogen/RGD peptide binding (ie, LIBSs) are the same as those induced by fiban drugs. Bougie and colleagues showed that LIBSs are different from mimetic-induced binding sites (MIBSs), a term they used for determinants induced by fiban, but not fibrinogen, binding to GPIIb/IIIa. They also found that FIT antibodies did not recognize LIBS determinants as previously expected but instead recognized conformational changes in GPIIb/IIIa induced by the fiban drugs (MIBSs). In most patients with eptifibatide-induced and tirofiban-induced thrombocytopenia, antibody binding was drug specific and occurred only when the integrin complexed with the sensitizing drug. However, antibody cross-reactivity occurred only in a minority of patients. These investigators also reported that the MIBSs resided in the head region of GPIIb/IIIa, near the RGD recognition site, probably on the β-propeller domain of GPIIb or the βA domain of GPIIIa ( Fig. 3 A ). More precise epitope mapping, particularly identification of GPIIb or GPIIIa sequences that mediate antibody binding, may have to await further studies.

Thrombocytopenia induced by GPIIb/IIIa inhibitors. ( A ) Fiban-induced DITP. A neoantigen emerges after drug binding. This is recognized by the antibody. ( B ) Abciximab-induced DITP. Naturally occurring nonpathogenic antibodies recognize the papain cleavage site at the C terminus of abciximab. Pathogenic antibodies may recognize the murine portion of the molecule. ( C ) Immune-complex or innocent-bystander mechanism. The immune complex binds to platelets via the Fc receptor and the platelet is destroyed/activated as an innocent bystander.

Abciximab-induced thrombocytopenia

Abciximab is a chimeric (mouse/human) monoclonal antibody (Fab fragment) that has specificity for the βA domain of GPIIIa. Because abciximab binds close to the RGD recognition site of GPIIIa, it blocks fibrinogen binding to activated GPIIb/IIIa and hence inhibits platelet thrombus formation. About 1% to 2% of patients given the drug develop acute, often severe, thrombocytopenia on first drug exposure, and 10% to 12% on the second and subsequent exposure. Like FIT, naturally occurring abciximab-dependent antibodies were detected in patients receiving the drug and in healthy volunteers never previously exposed to abciximab. There is a subgroup of patients who developed acute thrombocytopenia about a week after infusion of the drug. These patients presumably did not have naturally occurring antibodies and acquired the antibody on drug sensitization. The antibody reached a level that could react with the abciximab that remained on platelets approximately 7 days later.

Unlike other DITPs, both abciximab-dependent antibodies recognize the drug (abciximab) when bound to platelets. However, the patient antibody recognizes a mouse domain in the chimeric Fab. In contrast, the antibody in healthy individuals is specific for the papain cleavage site introduced during the manufacturing process and, consequently, binding can be inhibited by Fab fragments prepared from normal immunoglobulin (Ig) G (see Fig. 3 B).

How may drug-dependent antibodies cause thrombocytopenia?